Timing Analysis and Timing Predictability

tele-TASK-Aufzeichnung des Vortrags

Hard real-time systems need safe and tight upper bounds for their execution times to show that they satisfy their timing constraints. To determine such bounds is difficult because on modern processor architectures, the execution time of a machine instruction can vary several orders of magnitude. It strongly depends on the execution history.

In our approach to the determination of upper bounds on execution times, static program analysis is used to compute invariants about the set of all execution states at an instruction. These invariants are then used to derive safe and tight upper bounds on the execution time of the instruction in all these execution states. Based on these upper bounds a longest path through the program is determined.

The precision of the static analyses depends on system characteristics both of the underlying hardware platform and of the software layers. Strong influence on the variability of execution times and therefore on the predictability of the timing behaviour on the hardware side stems from properties of the memory hierarchy.

The talk will give an introduction into our method to determine upper execution-time bounds, which is used in the aeronautics and the automotive industry, introduces a notion of timing predictability, and it presents first quantitative results on the timing predictability of different cache architectures.

This is joint work with Christoph Berg, Daniel Grund, Jan Reineke.

Sprache:    Deutsch
Gastgeber: Prof. Dr. Christoph Meinel
Ort:           HS 2
Datum:      Donnerstag, 02. November 2006
Zeit:          16:00